Rovo Dev CLI meets the internet: An AI butler with planet-scale context

[#1]

Intro

This story starts, as all great engineering tales do… with pizza.

My team was planning a get‑together and my manager said, “Let’s order pizza for ~20 people.”
Easy, right? Except I’ve never ordered pizza for that many humans in my life.

Suddenly I’m juggling questions like:

• How many pizzas do hungry developers eat on average?

• Which nearby places get great reviews and can deliver by 5 p.m. without nuking the budget for swag?

• How do I keep vegans, vegetarians, meat lovers, and the “I don’t do cheese” crowd all happy?

I caught myself doom‑scrolling through menus, reviews, delivery estimates, and allergy notes and thought: why am I doing this?

Isn’t this exactly the type of chaotic, constraint‑filled decision that an AI agent should handle for me?

That little pizza crisis is what sparked the question behind this post:

What happens when Rovo Dev CLI meets the internet and turns into an AI butler with planet‑scale context?

Plugging the internet into Rovo Dev

We’ll use Chrome DevTools MCP, an open-source MCP that can control Chrome, to let Rovo Dev browse the internet.

Requirements (as of time of writing)

• Node v20.19 or newer

• Chrome current stable version

• npm

1. Open your Rovo Dev CLI MCP config

Rovo Dev CLI looks for MCP server configs in ~/.rovodev/mcp.json.

Open the file:

code ~/.rovodev/mcp.json
# or use your favourite editor

2. Add the Chrome DevTools MCP server

Add a chrome-devtools entry under mcpServers. For example:

{
  "mcpServers": {
    "chrome-devtools": {
      "command": "npx",
      "args": [
        "chrome-devtools-mcp@latest",
        "--isolated=true",
        "--no-performance-crux",
        "--no-usage-statistics"
      ]
    }
  }
}

If you already have other MCP servers configured, just merge this chrome-devtools block into your existing mcpServers object.

3. Run Rovo Dev CLI

Now start Rovo Dev CLI as usual:

acli rovodev run

Testing its capability

For this demo, I gave Rovo Dev CLI a simple but real task from my pizza crisis:

Prompt:
“Find me up‑to‑date top 5 pizza places near Wynyard Station that do deliveries.”

Rovo Dev CLI spins up a new Chrome instance and navigates the web to start its quest for pizza places.

Here’s a snippet of the result I got:

Here are the top 5 pizza places near Wynyard Station that offer delivery:

1 NAME WITHHELD PENDING LEGAL REVIEW
• Location: AAA York Lane, Sydney (Directly next to Wynyard Station)
• Rating: ~4.3/5
• Delivery: Yes, available via their website and major delivery apps.
• Style: Traditional Italian, known for authentic flavors.

2 LET'S PRETEND YOU DON'T RECOGNISE IT
• Location: BBB Jamison St (The Met Centre), Sydney (Very close to Wynyard)
• Rating: ~4.5/5
• Delivery: Yes, available on Uber Eats and for pickup.
• Style: Neapolitan-style wood-fired pizza, casual eatery.

.... More text

What’s important here is that Rovo Dev isn’t guessing or hallucinating: it’s actually browsing live web pages!

Amazing! Now let’s ask it to compile the order.

Prompt:
“Let's choose TOP SECRET PIZZA VENDOR 1. Can you compile me a list to order? I need to feed 20 hungry developers. One is vegan, one is vegetarian and one is lactose intolerant. The budget is $750.”

Here’s the result:

                                  The Order List

Dietary Specials (3 Pizzas)

• 1x Pizza Marinara (For Vegan)

• Tomato, garlic, oregano, extra virgin olive oil (No cheese).

• ~$26.00

• 1x Pizza Capricciosa - NO CHEESE (For Lactose Intolerant)

• Tomato, ham, artichokes, mushrooms, olives (Ask for NO mozzarella).

• ~$32.00

• 1x Pizza Norma (For Vegetarian)

• Tomato, eggplant, dry ricotta, basil.

• ~$30.00

General Group Sharing (13 Pizzas)

• 4x Pizza Margherita

• Tomato, fior di latte, basil. (Classic crowd pleaser)

• 4 x $28.00 = $112.00

…. more text ….

                                  Cost Breakdown

• Food Total: ~$667.00

• Remaining Budget: ~$83.00 (Perfect for delivery fees ~ $10-20 and driver tip).

From a single natural‑language prompt, the agent:

• Parsed a live menu from the web (so I didn’t have to squint at PDFs like it’s 2003)

• Respected dietary constraints

• Planned portions for ~20 hungry developers

• Stayed within the $750 budget and even left some buffer for delivery fees and guilt snacks

All I had to do was skim the list and nod like I did all the hard work.

Closing note

The possibilities are endless.

Some other productive errands you can throw at it:

• Find typos and alignment issues on your website and inject red boxes to make them obvious

• Check the performance of your website

• Compare products A, B, and C in terms of pricing and fit for purpose

• Identify any breaking changes between Go 1.25 and Go 1.26 (if any), and summarise the migration impact

I hope you enjoyed experimenting with it!

[Bernandina Ortiz]

@Albert Sundjaja, I really enjoyed this post. I need to show this to Alex Ortiz. He has been telling Atlassians that his dream feature is Rovo ordering and delivering pizza at Team ’26 Anaheim. 🍕

[Albert Sundjaja]

@Bernandina Ortiz Thank you! I believe it is entirely possible for Rovo to schedule pizza ordering and do it end-to-end without any human intervention with the right MCP :)

[Matthias Gaiser _K15t_]

Nice story - and thank you for the pointer to the Chrome DevTools MCP.

Here's an mcp project for ordering pizza via Domino's - sounds like the perfect Atlassian use case since they're featuring Domino's every now and then.

@Bernandina Ortiz & @Alex Ortiz - unfortunately I won't be in Anaheim this year, but I want to hear some stories how you made that possible.